Slashdot Mirror
New Linux TPC-H Record Set
prostoalex writes: "New TPC-H world record for performance and scalability of database software on Linux platform has been set. The winner - Oracle 10g running on a four-node Lenovo Cluster Server DeepComp 6800, each with four Intel Itanium 2 1.3 GHz processors. Oracle also emphasizes that it's 3.5 times more performance than similar IBM DB2 benchmark. TPC-H benchmarks are available at TPC Web site."
Now SCO's gonna want $2800 for a license.
"I might have made a tactical error in not going to a physician for 20 years." -- Warren Zevon
there are basically three type of clusters:
1) shared nothing: in this, each computer is only connected to each other via simple IP network. no disks are shared. each machine serves part of data. these cluster doesn't work reliably when you have to aggregations. e.g. if one of the machine fails and you try to to "avg()" and if the data is spread across machines, the query would fail, since one of the machine is not available. most enterprise apps cannot work in this config without degradation. e.g. IBM study showed that 2 node cluster is slower and less reliable than 1 node system when running SAP.
IBM on windows and unix and MS uses this type of clustering (also called federated database approach or shared nothing approach).
2) shared disk between two computers: in this case, there are multiple machines and multiple disks. each disk is atleast connected to two computers. if one of the computer fails, other takes over. no mainstream database uses this mode, but it is used by hp-nonstop. still, each machine serves up part of the data and hence standard enterprise apps like SAP etc cannot take clustering advantage without lot of modification.
3) shared everything: in this, each disk is connected to all the machines in the cluster. any number of machines can fail and yet the system would keep running as long as atleast one machine is up. this is used by Oracle. all the machine sees all the data. standard apps like SAP etc can be run in this kind of configs with minor modification or no modification at all. this method is also used by IBM in their mainframe database (which outsells their windows and unix database by huge margine). most enterprise apps are deployed in this type of cluster configuration.
the approach one is simpler from hardware point of view. also, for database kernel writers, this is the easiest to implement. however, the user would need to break up data judiciously and spread acros s machines. also adding a node and removing a node will require re-partitioning of data. mostly only custom apps which are fully aware of your partitioning etc will be able to take advantage.
it is also easy to make it scale for simple custom app and so most of TPC-C benchmarks are published in this configuration.
approach 3 requires special shared disk system. the database implementation is very complex. the kernel writers have to worry about two computers simultaneously accessing disks or overwriting each others data etc. this is the thing that Oracle is pushing across all platforms and IBM is pushing for its mainframes.
approach 2 is similar to approach 1 except that it adds redundancy and hence is more reliable.
Shouldn't that read "New TPC-H Record Set Using Oracle?"
The article didn't give much details, but how much of this performance is directly attributable to Linux (specifically Red Hat AS3)? What was the OS of the system it beat? Could that also have been Linux? How much of the performance can be attributed to the (suspiciously un-Beowulf) Lenovo cluster?
From what I know of benchmarks, the numbers given reflect real-world preformance, to within one order of magnitude.
At first, I thought, It's just a press release, big deal... But wait, they used Linux, so it must be another straw on the back of the camel knows as the Closed Source Business Model. But wait, it's running Oracle, so it must therefore be evil. Aieeeeeeeeeeeeee!!!!
Holy crap this story is useless! Go to the TCP-H site and actually look at the results, it really is nothing even remotely impressive.
- It's NOT the fastest TCP-H result, it's the fastest LINUX TCP-H 1000GB result. Actually it's the ONLY Linux TCP-H 1000GB result. 5th of 8 overall
- It's not even offering very good bang for your buck, coming in 5th of 8 for Price/QphH ($156 US according to today's currency exchange). The only systems it managed to beat are two outdated systems (both from HP) and an old price for a Fujitsu system, quoted in euro (the same system offers the same performance but a lower price on a newer entry quoted in US $).
In short, if anything this suggests that Linux is a BAD choice for this work! The performance isn't there and the cost is high.
Where things get REALLY bad though is the claim that this is "3.5 times faster" than a system running IBM's DB2. This is just 100% pure bullshit! The new Linux/Oracle system runs 1.3GHz Itanium2 processors and Oracle 10g. The HP/Windows/DB2 system runs 900MHz Xeon processors and runs DB2 7.2 (8.1 is current version). What's more, the Oracle/Linux system isn't even 3.5 times faster, it's just 3.5 times faster PER PROCESSOR! Great, your brand-spanking new Itanium2 is 3.5 times faster than four year old Xeon 900MHz chips. Whoopie!
Note: if you do want to see impressive Linux results, look at what IBM is doing with their Opteron cluster and DB2 running under SuSE Linux. They turned in the top results in the two TPC-H tests they entered (100GB and 300GB).
I was wondering who bought all the Itanium 2s....
The OS is largely irrelevant to speed tests which never swap or do I/O, like generating graphics. But servers show weaknesses in an OS like nothing else, since they really hammer context switches and I/O.
This IS significant. It shows the suits that Linux can handle swap intensive tasks, even tho they don't know that is what it shows.
Infuriate left and right